Participants and procedure

The participants’ recruitment process was conducted in collaboration with the disability care services of the Spanish universities. These services have access to most of the students with disabilities enrolled at each institution, since they provide these students with the support needed in order to adjust to the academic requirements and the university lifestyle. According to an acknowledged institutional guide on disability care at Spanish universities (^{Fundación Universia, 2016}), there was a population of 20,695 university students with disabilities during the data collection period (fall 2016 fall 2017). An accessible population of 15,038 students with disabilities was estimated from the data provided by the 55 universities that were involved in the data collection process. Drawing on ^{Cochran’s (1977)} sampling techniques, it was determined that 997 participants were needed for a statistically valid sample size (Confidence Level = 95%; Population proportion = 50%; Margin of error = 3%). Given that the data protection policies of the different universities prevented direct access to students, the disability care services sent the digital survey by institutional email to the accessible students.

A total of 1264 university students with disabilities completed the survey. After excluding respondents with missing data, 1076 participants (530 men, 546 women) remained for the analyses. Table 1 shows participants’ sociodemographic characteristics. All participants gave their informed consent to be included in the study, and the procedures and materials used were approved by the Ethics Committee of the University of Valencia.

Table 1 Socio-demographic characteristics of the sample and accomplishment of the World Health Organization physical activity recommendations by each characteristic (n= 1076). 

Measures

The TPB constructs were measured using a questionnaire, which was developed drawing on previous work by ^{Latimer and Martin Ginis (2005)} and validated using a sample of Spanish university students with disabilities by ^{Úbeda-Colomer, Pérez-Samaniego, and Devís-Devís (2018)}. Consistent with other evidence in the PA domain (^{Armitage & Conner, 1999}; ^{Terry & O’Leary, 1995}), the validation process of the Spanish questionnaire supported splitting up the PBC construct into two independent factors: self-efficacy and controllability. The results of the validation process in terms of adjustment were excellent (CFI = .97; RMSEA = .057; IC90% = .049 .066) and the Alpha coefficients for the subscales ranged from .74 to .93, thus showing good internal consistency.

Attitudes. Attitudes were assessed using four adjective-pairs that measured both the instrumental (bad-good, worthless-valuable) and the experiential (stressful-relaxing, boring-funny) component of attitudes. Items were preceded by the statement, “I think that participating in physical activity for at least 30 minutes on at least three days a week would be…”.

Subjective norms. Subjective norms were assessed with two items and the common stem “Most people who are important to me”: (a) “think I should participate in physical activity for at least 30 minutes on at least three days a week”, and (b) “approve of me participating in physical activity for at least 30 minutes on at least three days a week”.

Self-efficacy. Self-efficacy was measured with two items: (1) “How confident are you that you will be able to participate in physical activity for at least 30 minutes on at least three days a week?”; and (2) “To what extent do you see yourself as being capable of participating in physical activity for at least 30 minutes on at least three days a week?”.

Controllability. Controllability was measured with three items: (1) “How much personal control do you feel you have over whether you participate in physical activity for at least 30 minutes on at least three days a week?”; (2) “Whether or not I participate in physical activity for at least 30 minutes on at least three days a week is entirely up to me”; and (3) “How much do you feel that whether you participate in physical activity for at least 30 minutes on at least three days a week is beyond your control?”

Intentions. Intentions were assessed with two items: (1) “I will try to do at least 30 minutes of physical activity on at least three days a week”, and (2) “I intend to do at least 30 minutes of physical activity on at least three days a week”. All items were rated on 7-point Likert-type scales, with lower scores indicating more negative thoughts and feelings. The overall score for each construct was calculated as the mean of all the items of that construct.

Moreover, the reduced Spanish version of the Barriers to Physical Activity for People with Mobility Impairments (BPAQ-MI) was used to measure barriers to PA experienced by the participants. Drawing on the questionnaire developed by ^{Vasudevan et al. (2015)}, ^{Úbeda-Colomer, Peiró-Velert, and Devís-Devís (2018)} conducted a process of reduction, translation and validation of the instrument for the Spanish context using a sample of university students with disabilities. The results of the validation process in terms of adjustment were excellent (CFI = .97; RMSEA = .064; IC90% = .061, .067) and the Alpha coefficients for the different subscales ranged from .74 to .91, thus evidencing good internal consistency. With a total of 29 items, this instrument allows to equitably measure barriers across the four social ecological levels (intrapersonal, interpersonal, organizational and community).

Intrapersonal level. The intrapersonal level of barriers was measured with 7 items: “You get tired or fatigued”, “You were in pain”, “You were afraid of getting injured while being physically active”, “You lack the motivation to be physically active”, “You don’t have confidence in your ability to be physically active”, “You were embarrassed about your appearance while being physically active” and “You don’t see a reason to be physically fit”.

Interpersonal level. The interpersonal level of barriers was measured with 7 items: “Your friends didn’t assist you to be physically active”, “Your friends are not physically active”, “Your friends were not encouraging or supportive of your efforts to be physically active”, “Your family did not assist you to be physically active”, “Your family members are not physically active”, “Your family members were not encouraging or supportive of your efforts to be physically active” and “Your family did not think physical activity would be helpful to improve your health”.

Organizational level. The organizational level of barriers was measured with eight items: “Lack of adapted equipment/material at fitness centre”, “Lack of accessible showers/bathrooms/locker rooms at fitness centre”, “Lack of adaptation of fitness centre facilities (corridors, doors, elevators, etc.)”, “The economic cost was too high.”, “Lack of inclusive marketing at fitness centre”, “Lack of adapted programs or activities at fitness centre”, “Lack of adaptation of outdoor spaces (parks, ways, etc.)” and “Lack of assistance or training of the fitness centre staff”.

Community level. The community level was measured with 7 items: “Inaccessible sidewalks (gaps, lack of ramps, too narrow…).”, “Potholes in the streets, driveways or parking lots”, “The crosswalks lack traffic lights or they are not adapted (e.g. no sound when it is green…)”, “Lack of adapted transport to go to fitness centre”, “Lack of support staff to help you to go to fitness centre”, “The city traffic is dangerous for you.” and “The traffic lights or crosswalk signals change too quickly”.

All the items were presented in a matrix with a common statement at the top: “Please, think about the main barriers that hindered or prevented you from PA engagement during the last months. Then, rate each of the next barriers from 0 to 4, with 0 meaning ‘It has not been a barrier for me’ and 4 meaning ‘It has been a very important barrier for me’.

Data analyses

Homogeneity of variances and normality of all the variables were assessed using the Levene’s test and the Shapiro-Wilk test. Due to violation of the normality assumption, possible gender differences in the TPB constructs were tested using Mann-Whitney U tests, with r coefficients as a measure of the effect size (^{Rosenthal, 1991}). Finally, in order to examine which social ecological levels of barriers were predictors of selfefficacy and controllability, forward stepwise multiple regressions were conducted. This method was chosen instead of hierarchical regressions given the lack of evidence to determine the order of entry of the different social ecological levels in the model. Since preliminary analyses conducted showed no significant differences in the TPB constructs by type of disability, congenital-acquired disability and age, there was no need to control for these variables. The α level was set at p < 0.05 for all the analyses and multiple testing was accounted for using Bonferroni correction. All analyses were conducted using The Statistical Package for the Social Sciences for Windows (version 22.0; SPSS Inc., Chicago, IL).

Discussion

The present study is the first to analyse gender differences in theory-based predictors of PA in a sample of university students with different disabilities. It is also the first study examining if the social ecological levels of barriers to PA differently predict self-efficacy and controllability according to gender. The results obtained could thus be of great relevance for the sports services and the disability care services of universities in order to implement specific PA behaviour change interventions which differentially target men and women.

The behavioural control variables (self-efficacy and controllability) obtained the lowest scores, both in men and women. These results make sense in light of several review studies showing the wide range of barriers that people with disabilities face when trying to be physically active (e.g., ^{Jaarsma et al., 2014}; ^{Kissow, 2015}; ^{Martin, 2013}; ^{Martin Ginis et al., 2016}). In addition, women reported even lower PBC than men, which coincides with the results obtained by ^{Stapleton and Martin Ginis (2014)} in a sample of people with spinal cord injury. In this respect, according to the theory, past behaviour is a key determinant of self-efficacy. Therefore, the lower levels of PA reported by women with disabilities in several studies (e.g., ^{Úbeda Colomer et al., 2019}; ^{Valis & González, 2017}) might explain the lower scores on the PBC construct, thus generating a vicious circle. In addition, since the PBC construct is highly related to the absence or presence of barriers to perform the behaviour, these results are also consistent with some studies showing that women experience more barriers to PA than men (e.g., ^{Rimmer et al., 2000}; ^{Úbeda-Colomer, DevísDevís, & Sit, 2019}).

It is important to highlight that, according to ^{Ajzen (2002)}, PBC moderates the relationship between intentions and behaviour, so that individuals with higher behavioural control are more likely to translate intentions into actual behaviour. Given the low scores reported by the participants of this study on behavioural control measures, especially in the case of women, future interventions addressed to university students with disabilities would benefit from including action planning and coping strategies. These strategies could be useful in enhancing PBC and overcoming barriers to PA and have been proven effective to increase PA behaviour (^{Arbour-Nicitopoulos, Martin-Ginis, & Latimer, 2009}; ^{Carraro, & Gaudreau, 2013}). In addition, addressing the most important factors preventing PA engagement of university students with disabilities would increase their perceived control over the behaviour as well. For instance, offering affordable and adapted activities at university campuses would be of great importance, since the lack of adapted PA programs and their economic cost have been proven to be relevant barriers for this population (^{ÚbedaColomer et al., 2019}).

Given the gender differences observed in self-efficacy and controllability, the secondary aim of this study was to determine if the social ecological levels of barriers differently predicted these constructs for men and women. Intrapersonal barriers were the only significant predictor of self-efficacy for both men and women. In this regard, since lack of motivation has been identified as a relevant intrapersonal barrier to PA in university students with disabilities (^{Úbeda-Colomer et al., 2019}), interventions aiming to encourage physically active lifestyles in this population would benefit from including strategies such as self-monitoring or goal setting (^{Knittle et al., 2018}).

Regarding controllability, some gender differences were observed. While intrapersonal barriers predicted controllability both for men and women, organizational barriers were an additional predictor for men, while for women, community barriers were an additional predictor. It is important to note that these results do not mean that organizational barriers would not affect women, but rather that women participating in this study would not have enough experience participating in PA to realize about organizational barriers. That is, women seem to not experience organizational barriers because they are not overcoming the barriers faced at the intrapersonal level. Even though women are willing to engage in PA, a lack of confidence in their ability to do so would prevent them from trying. Men have higher self-efficacy, so they are more likely to translate PA intentions into PA behaviour according to ^{Ajzen (2002)}. Therefore, once they try to be physically active, they would realize the presence of barriers at the organizational level, such as a lack of adapted PA programs or a lack of training of the staff at the fitness centre (^{Úbeda-Colomer et al., 2019}). Instead, since women report lower PBC, they would translate intentions into behaviour to a lesser extent than men, so they would be less aware of the barriers that could appear at the organizational level.

In this respect, the social ecological theory argues that the different levels are interconnected and can exert reciprocal influence on each other. This is especially relevant to consider when implementing policies for the promotion of physically active lifestyles among people with disabilities, given the wide range of barriers that this population face at all the levels (^{Martin Ginis et al., 2016}). For instance, interventions at the community level, such as improving accessibility of the city, could have a positive impact on women’s self-efficacy, thus making them more likely to engage in PA. Overall, the results of the present paper show the importance of adopting multi-level approaches towards health and PA promotion in people with disabilities. Since behavioural control and barriers at different social ecological levels have been revealed as important when it comes to PA in university students with disabilities, interventionists seeking to increase PA in this population should pay due attention to these factors, rather than solely focusing on the individual level. To do so, interdisciplinary research and practice that considers disability and health as multifaceted constructs is needed (^{Agiovlasitis, Yun, Jin, McCubbin, & Motl, 2018}).